Sovereign tax measures: hostile act or ordinary sovereignty

Tax is the best protected of all sovereign powers, and that protection does not come from an oversight: it is written down. Bilateral investment protection treaties almost always reserve tax matters, and political risk policies explicitly exclude taxation that is in good faith, non-discriminatory and of general application. A state wishing to take value without triggering the covers therefore has an instrument designated in advance, whose use is presumed legitimate and whose challenge is framed by its own courts. This is the hardest version of the problem seen with creeping expropriation: there, the normal exercise of regulatory power was a defense argument; here, it is a written exclusion.

The exclusion must therefore be read for what it actually says, because it is never absolute. It is qualified, and the qualifiers are the entire battleground: in good faith, non-discriminatory, of general application. A levy can be so confiscatory, so retroactive or so narrowly targeted that it ceases in substance to be a tax and becomes a taking dressed as one. Four markers serve to establish this, and they accumulate rather than being chosen among: retroactive application to closed financial years, a rate or a base exceeding the economic capacity of the activity to the point of making it unviable, a base shaped so as to reach in practice only foreign holders, and the gap between the measure and the procedure the state gave itself. None of these markers suffices alone; three together move a file.

Alongside political cover sits a distinct product that must be known so as not to be confused with it, tax liability insurance. Its peculiarity is that it does not cover a contingency in the usual sense: it covers an identified, quantified position that the insured holds to be correct and fears an authority will challenge. The insurer therefore models no loss distribution, it prices a legal opinion, assessing the strength of a tax analysis and the probability that an inspector will prefer another. It is an unusual operation in insurance, and it explains the shape of the contract: the amount is known in advance, the question is binary, and the premium reflects a forecast about a reading of text rather than about an event.

That product immediately raises a moral hazard objection that must be put plainly: if the rejection of a tax position can be insured, aggressive positions can be taken and their cost left to the insurer. The market answers with structural mechanisms rather than with trust. Underwriting requires a reasoned opinion from counsel, at a stated level of conviction, and that opinion is the basis of the rate. Disclosure must be complete, and concealing a known weakness empties the cover. Positions with no reasonable basis are excluded, as is the situation of a taxpayer who knew an audit had begun. But the rule holding the whole structure up is simpler than all the others: the insured position must have been formed before cover was sought. A position invented in order to be covered is not insurable, because insurance must not alter the behavior it covers, and the entire apparatus aims at nothing else.

A difficulty peculiar to tax ground then delays everyone. Treaties that reserve tax matters frequently require, before an expropriation claim founded on a tax measure can proceed, that the tax authorities of both states be seized first, or that local remedies be exhausted. A state taking through tax therefore buys itself, by construction, several years of delay, during which the investor bears the measure. To this must be added a very widespread and rarely anticipated procedural rule: an appeal is admissible only after payment of all or part of the reassessment. The investor thus funds the taking it disputes, and the cash leaves before any recovery is conceivable, which makes liquidity a subject distinct from the merits and one that must be handled separately.

That leaves the difficulty that decides files, and it lies in the investor pleading simultaneously before two forums that do not expect the same account. Before the tax authority and the tax judge, it argues that its returns complied with the country's tax law and that its reading was the right one, which treats the system as legitimate and competent. Before an arbitral tribunal, it argues that this same system was used as an instrument of dispossession, which requires saying it did not operate as a tax system. Both cases are defensible, and they are not necessarily contradictory, but each weakens the other as soon as one is drafted without thinking of the other. A tax submission that concedes, in order to win a technical point, that the disputed circular is a legitimate interpretive act will be produced as it stands before the arbitral tribunal, and will cost more there than it earned.

Practical conduct follows from these constraints and turns on ordinary documents. The contemporaneous tax file must be kept: rulings obtained, exchanges with the authority, the reasoning adopted at the time the returns were prepared, since this is what demonstrates the good faith of the position and deprives the state of its most convenient argument. And one must gather, while it is still possible, how comparable domestic taxpayers were treated on the same question. That is the proof of discrimination, meaning the strongest marker available, and it is information that disappears: once comparable reassessments are settled, compromised or time-barred, nobody can any longer establish that the new rule was applied to one taxpayer alone.